文献类型：期刊文献

中文题名：一种基于改进SURF和K-Means聚类的布料图像匹配算法

英文题名：A Fabric Image Matching Algorithm Based on Improved SURF and K-Means Clustering

作者：张雪芹[1];刘远远[1];曹逸尘[1];张鹏飞[1]

机构：[1]华东理工大学信息科学与工程学院,上海200237

年份：2017

卷号：43

期号：1

起止页码：105

中文期刊名：华东理工大学学报（自然科学版）

外文期刊名：Journal of East China University of Science and Technology

收录：CSTPCD;;Scopus;北大核心:【北大核心2014】；CSCD:【CSCD_E2017_2018】；

基金：国家自然科学基金(61371150)

语种：中文

中文关键词：布料图像匹配;SURF特征;小波变换;K-Means聚类

外文关键词：fabr ic image matching; SURF feature; wavelet trans form; K-Means clustering

摘要：计算机图像智能处理技术为服装设计师开展设计、启发灵感提供了方便和可能。通过提取布料图像的SURF特征可以实现布料图像形状分析,但由于SURF特征维数高、特征提取是基于灰度图进行,因此存在匹配速度慢、匹配结果不够符合人眼视觉特点的问题。本文提出了基于小波变换的自适应SURF特征提取算法和基于K-Means聚类的布料图像颜色分析方法。通过融合图像形状特征、颜色特征,加快了布料图像匹配速度,使布料图像的匹配结果更加符合人眼视觉感受。在8种不同类型布料图像上的实验验证了该算法的有效性。
Computer intel l igent image processing technology can provide an effective aid for dress designer. By extracting the SURF features,the image shape of the cloth can be recognized. However,due to the high feature dimension and the grayscale based feature extraction method of SURF, there exist shortcomings, e. g, slow image matching speed and the matching result is not enough to match the characteristics of human visual. Hence,this paper proposes an adaptive SURF feature extraction algorithm based on wavelet transform and an image color analysis method based on K-Means clustering. By fusing the shape and color feature of the image,the matching speed is accelerated and the matching results are made more accord with the human visual perception. Experiments via 8 different kinds of fabric images show the effectiveness of the proposed algorithm.